# Measuring RF Filters with a Homebrew 1.5GHz Scalar Network Analyser Costing £32/\$48

I already have useful low/medium frequency signal sources, spectrum analysers and oscilloscopes. Now I want to inexpensively measure RF filters and transmission line imperfections. This is possible for only £32/\$48, as illustrated by the measured response of 300MHz and 460MHz high-pass filters and an open-stub transmission line filter:

Traditionally such measurements have required an RF sine-wave tracking signal generator plus spectrum analyser, but they are expensive and/or bulky, and their frequencies have to be synchronised. The frequency and amplitude also have to be calibrated, which is a “Catch-22” if you don’t already have the test equipment. All those disadvantages can be avoided by using the noise source and RTL SDR dongle mentioned here.  However, they bring their own limitations:

• frequency domain artefacts discussed here
• since power is simultaneously transmitted at all frequencies, it is relatively easy to saturate the receiver and cause non-linear components to generate harmonics
• amplitude response only, not phase response
• limited dynamic range, and slow measurement times

Nonetheless, despite those limitations, useful results can be obtained.

### Technique

The measurement is made in these stages:

1. connect the noise source to the receiver and record the power measured by the receiver, $P_{cal}(f)$. To avoid saturating the receiver or network-under-test, it may be necessary to insert an attenuator pad
2. insert the network being measured, and record the power measured by the receiver, $P_{raw}(f)$.
3. the network’s response is then $N(f) = P_{raw}(f) - P_{cal}(f)$

### Results

The transmission line filter is 1.02m long, and open-circuit at the end. The 200MHz HP filter is a tee-filter with 10pF capacitors and a 24nH inductor. The 460MHz HP filter is a tee-filter with 4.7pF capacitors and an 11nH inductor. All components are 0603 SMT devices mounted on a small CPWG board with SMA connectors.

With a 10dB pad inserted between the noise source and receiver, the $P_{cal}(f)$ (blue) and $P_{raw}(f)$ (red, green) lines show the system’s sensitivity falling above 300MHz, the noise floor at about -29dB, the dynamic range being around 35dB:

Hence the calculated filter responses, $N(f)$ are:

The open-stub filter’s troughs do not rise with frequency; that is an artefact of the receivers falling sensitivity. The high-pass filter’s LF response does not level off at -30dB; that level and associated artefacts are a consequence of the receiver’s limited dynamic range.